Measuring benefits from reduced eutrophication

model which allocated abatement tasks among the Baltic countries so as to jointly meet the 50 target, at least cost Gren et al., 1995. Investment costs were annualized at 7 discount rate and averaged over the 20-year period corresponding to the JCP time horizon.

4. Measuring benefits from reduced eutrophication

Estimation of benefits that may be attributed to pollution abatement in the region seems to be an even more troublesome task than measuring the costs of pollution abatement. The CVM is consid- ered to be the only economic method of natural resource valuation capable of capturing non-use value components. The methodology of CVM has improved significantly over approximately 30 years of its development. Nevertheless the method itself, its validity and reliability, are still a source of controversy. Moreover, as evident from the recent discussion in Ecological Economics on the global environment valuation exercise Costanza et al., 1998, there are still unresolved problems even more fundamental than just valuation techniques. According to Whitehead et al. 1995 technical reliability of WTP estimations is defined in terms of the extent to which the WTP measure is im- plied by nonrandom sources and the stability of the measure over time. Validity is the extent that the measured value corresponds to the theoretical definition of value. Different biases may affect both validity and reliability of the WTP estimates. Such biases were tested in some experiments: see, e.g. Hanemann 1994 for well-known experi- ments with strawberries; Cummings and Harrison 1994 for a test on the strategic bias; Niewijk 1994 for an embedding bias. It is difficult to draw one conclusion based on these experi- ments — some studies tend to praise CVM while others deny its reliability. In spite of the controversies regarding biases, the values produced in CVM experiments have been accepted by many researchers as well as by some policy-makers. One of the most recent hand- books devoted to environmental economics states: ‘‘The WTP … measures of economic value can be used as restrictions to guide policy or can be included, with caution, in the bottom-line cost – benefit analysis used to support public policy’’ Hanley et al., 1997. In the US, CVM has been adopted for practical use in several instances. The Department of Inte- rior accepted CVM as the best available proce- dure for assessing monetary damages to resources. This decision was challenged by the State of Ohio in a Federal District Court in 1986, but the court affirmed it Cummings and Harrison, 1994. In 1994, National Oceanic and Atmospheric Administration NOAA recommended using CVM in oil spills damage assessments. NOAA’s decision has been taken after a panel of scientists had analyzed the usefulness and reliability of CVM in evaluating non-use values. Also EPA has begun to use CVM studies in cost – benefit evalua- tions of its regulations Niewijk, 1994. The NOAA panel report Arrow et al., 1993 sets guidelines for researchers who undertake CVM studies. Compliance with its requirements is likely to alleviate biases, leading to higher reliabil- ity. It should be noted here that the CVM surveys described in the latter part of the paper conform to the rules set in the NOAA panel report. Three Baltic countries, Poland, Lithuania and Sweden, have been the subjects of empirical stud- ies aimed at eliciting aggregate national WTP for improvement of the quality of the Baltic Sea. The researchers’ additional task then was to extrapo- late the results of empirical valuation from these Table 2 Annual costs of reducing the nutrient load to the Baltic Sea by 50 a Reduction Cost −g i p i 10 6 Finland 380 52 42 Sweden 710 51 390 Denmark 530 39 Germany 1280 63 Poland Lithuania 330 55 56 Latvia 240 Estonia 55 200 Russia 80 44 50 4140 Total a Source: Gren et al. 1995. three countries to the whole Baltic region using a benefit transfer approach. According to Bergland et al. 1995 there are generally three types of benefit transfer approaches: 1 transferring mean unit values; 2 transferring adjusted unit values; and 3 transferring demand functions. Often benefit transfer methods are not a reliable tool. It is not the purpose of this section to test reliability and validity of benefit transfers, al- though the empirical data collected provide quite a good starting point for a future work on this subject. What we try to achieve here is to compile empirical CVM results in such a way that both methodological discrepancies and country-specific factors may be accounted for as much as possible. We will present findings obtained from five empirical CVM studies on country-wide samples: three from Poland, one from Sweden and one from Lithuania. All studies deal with the same problem, i.e. measuring benefits from reduced eu- trophication of the Baltic Sea. The three Polish studies vary in methodology — the first, prelimi- nary, survey employs an open-ended OE WTP question, while the other two studies use a di- chotomous choice DC type of question. The last of the Polish studies was carried out as a mail survey, while the first two were carried out as face-to-face interviews. Yet another CVM study was carried out at the Polish Baltic Sea coast see Z : ylicz et al. 1995 for the results. However, being aware of a bias resulting from a specific type of the sample beach users, we will skip this one in our analysis. 4 . 1 . Analysis of empirical results The two surveys presented in this paper are fully comparable. The Lithuanian survey carried out on 10 – 18 October 1994 was based on the earlier Polish pilot study implemented on 21 – 28 July of the same year. Both surveys were carried out by professional polling agencies of the respec- tive countries on representative, country-wide samples. The Polish sample consisted of 1166 respondents and the Lithuanian one of 1002. The Polish pilot sur6ey was the first in the series of CVM experiments. It was designed as a preparatory work before other surveys. The main purpose of the pilot study was to check the ade- quacy of the valuation scenario and questions, as well as to estimate the range of initial bids to be used in the next surveys. The questionnaire was short and consisted of a scenario and three questions. The question about support for a ‘Baltic tax’ was answered positively by 60 of respondents and negatively by 39.2. A mere 0.9 of the respondents stated that it was difficult to say if they would or would not support the tax ten cases. Those who gave a positive answer to the first question or said they did not know were subsequently asked about their opinion on an appropriate value of the Baltic tax an open- ended question. Those who would not support the introduction of such a tax were asked about the reasons of refusal. How to estimate mean WTP in situations where a significant number of the respondents do not answer the valuation question? Different authors take different approaches here: some e.g. Lang- ford and Bateman, 1993 tend to focus on positive respondents only while others e.g. Gren et al., 1995 assume zero value for non-respondents. An interesting approach was proposed by Berg- land et al. 1995. Respondents who rejected the idea of payment are divided into two groups: ‘legitimate zero bidders’, who did not value the resource in question or could not afford to pay, and ‘protest bidders’, who viewed the idea of payments as not being their responsibility. We have followed this approach by making a distinc- tion between protest and zero bidders according to different reasons for payment refusal. Fig. 1 shows the breakdown of reasons for rejecting the idea of the Baltic tax. In the group of zero bidders we decided to place those who can- not afford to pay as well as those who think that there are other, more important goals to be financed. Assigning zero values to these categories does not necessarily mean that those respondents do not value the Baltic Sea or its recovery. Their WTP in the situation described in the valuation scenario, however, can be considered zero. The reasons such as ‘‘I pay too much taxes already’’, ‘‘I do not believe that my contribution may lead to any improvement’’ and ‘‘I am not As can be seen from Fig. 2, the option ‘‘I cannot afford to pay’’ has been chosen less fre- quently relative to the Polish study. The mean and median were calculated as WTP Lithuania = 27.92 LIT : 7 at exchange rate 4 LIT, M Lithuania = 10 LIT : 2.5; N = 697. Regression models re- vealed no significant factors except for household income. Except for the Polish pilot and Lithuanian ex- periments, all other Baltic CVM studies employ a DC-type of WTP question, which is preferred by many authors as more natural and resembling situation at real markets. Below the results of two other Polish CVM studies are summarized for comparison and discussion of the most reliable, approximate value of WTP for the Baltic Sea recovery. The main Polish study was carried out on 25 – 29 November 1994 on a representative all-Poland sample of 1162 respondents, as a part of a larger sociological survey. Valuation scenario gave a short description of eutrophication effects ob- Fig. 1. Zero bidders exploded versus protest bidders in the Polish pilot study. 1 I cannot afford to pay but I would do so otherwise; 2 I think I pay too much in taxes already; 3 I do not believe that my contribution will lead to any improve- ment; 4 I do not think that this is the most important problem in our country; there are other, more important goals; 5 I do not feel responsible for the condition of the Baltic Sea; let the responsible parties pay; 6 other reasons. responsible for the state of the Baltic Sea, let others, who are responsible, pay’’, indicate that respondents have a particular reason for refusal, not connected with the value of payment. This is why we decided to exclude this group of respon- dents from WTP calculation Fig. 1. The mean value of WTP Pilot including zero bidders simply as zero values and protest bidders as missing equals 31.98 PLN : 14 at the ex- change rate of 2.32 PLN prevailing at the time of the survey, N = 820. Median value is 10 PLN : 4.3. Let us compare these results with the results from the Lithuanian CVM study, which applied the same scenario and questions. The only differ- ence was that the list of reasons for refusal in the Lithuanian study contains an additional option: ‘‘I am not able to value clean Baltic Sea’’ for the full menu of other options, see the footnote to the Fig. 2. The results of the Lithuanian survey are as follows. A total of 44.2 of the sample positively answered the question about support for the Baltic tax, while 41.6 of respondents said that they would not support the proposed action. A relatively high percentage of the sample, com- pared to the Polish study, did not know if they would or would not support a tax levied on all Lithuanians for the sake of protection of the Fig. 2. Zero bidders exploded versus protest bidders in the Lithuanian study. 1 I cannot afford to pay but I would do so otherwise; 2 I think I pay too much in taxes already; 3 I do not believe that my contribution will lead to any improvement; 4 I do not think that this is the most important problem in our country; there are other, more important goals; 5 I do not feel responsible for the condition of the Baltic Sea; let the responsible parties pay; 6 other reasons; 7 I am not able to value clean Baltic Sea. Fig. 3. Percentages of ‘yes’ and ‘no’ answers to the initial bid in the main Polish study. For DC responses the mean WTP is given by the area under the cumulative probability bid function. This area may be truncated at zero andor the value of the maximum initial bid. Here we have used a simple, not-truncated model. By allowing negative amounts we admit that we can- not exclude that someone who refused to pay anything would rather withdraw the funds that are already being spent by the economy on pollu- tion abatement in the Baltic Sea. By not truncat- ing the model at the maximum bid value, we take into account that the value accepted by a given respondent is not the maximum value this respon- dent might have accepted. For further discussion of this problem, see Langford and Bateman 1993. In the linear logistic model integrated over the entire set of real numbers the estimate of mean WTP is the same as the median, since the density function is symmetrical about its mean value. By substituting the definition of a median P No = P Yes = 0.5 into the linear logit model we can calculate the median as − ab. Thus in this special case also the arithmetic mean, WTP = − ab Based on the model coefficients we got the average value 169.33 PLN : 73 for positive bidders 713. According to the same criteria as before, 224 respondents were classified as zero bidders, and 225 as protest bidders. In order to modify the meanmedian value by including zero bidders we have multiplied WTP achieved from the model by the factor 713937 the ratio of positive bidders to positive and zero bidders. Hence WTP Main = 129 PLN : 55.5. Yet another CVM study was carried out in Poland in MayApril 1995. This time mail sur6eys were used to compare the results with Swedish studies which employed that method. Unfortu- nately, the method is characterized by a relatively low response rate in Poland. Moreover, overesti- mation may occur resulting from the fact that those who respond are usually more interested in the subject. In Poland the questionnaires with a Baltic valu- ation scenario were sent to 600 randomly chosen addresses; 304 have been filled out and returned, served in the Baltic Sea and mentioned a possibil- ity of carrying out an international clean-up ac- tion over the next 10 years. Such an action would call for collecting financial resources in each Baltic country in the form of an earmarked tax. The respondents were then asked if they would support such an action. A total of 62.5 of respondents said ‘‘yes’’, and 29.8 said ‘‘no’’; 6.7 said they did not know. A DC question on acceptancerejection of an initial bid followed. Initial bids and the ratio of acceptancerejection grouped according to initial bids that were randomly distributed across the sample are displayed in Fig. 3. Numbers displayed in bars represent percentage of yesno answers in groups of respondents that were given a particular bid. As expected, the percentage of acceptance falls as initial bids rise. For estimation of mean and median values of WTP, a linear logit model was used. The model with one explanatory variable — an initial bid value — has the following form: lnP No P Yes = a + b·BID where P No and P Yes represent the probability of refusal and acceptance of a given initial bid BID. The bid level revealed to be highly significant lnP No P Yes = − 1.1345 + 0.0067 BID see Appendix A for statistical details of the equation. Table 3 CVM studies results: mean WTP values in 1995 US including zero bidders and excluding protest bidders Country Lithuania Poland Sweden Pilot OE Main DC Type of empirical study Mail DC Pilot OE Mail DC WTP 14 7 56 102 458 giving the response rate of 50.7. The majority of respondents 54.3 said that they would be willing to contribute part of their income in the form of a special tax for the sake of protection of the Baltic Sea. A linear logit model was esti- mated as: lnP No P Yes = − 1.7787 + 0.0037 BID see Appendix A for statistical details of the equation. From this we obtain mean WTP = − ab = 480.73 PLN for positive bidders only. Including zero bidders we obtained the value of WTP Mail = 236.32 PLN : 102. The Swedish mail sur6ey with the same DC type of WTP question was carried out on a 700- respondent sample response rate 60. The mean annual WTP reached 5800 SEK when only positive bidders were taken into account Gren et al., 1995. Using the method of distinguishing zero bidders and protest bidders based on So¨derqvist, 1998 we estimated the mean WTP Sweden at 3439 SEK which is equivalent to 458 at the exchange rate of 7.5 SEK. 4 . 2 . WTP estimates and benefit transfers Table 3 summarizes WTP values obtained from various studies. In the subsequent part of the paper these values will be used to estimate the benefits ratios: proportions of each country’s marginal WTP to the sum of marginal WTPs p i p N , and then plugged into the Chander – Tulkens formula. As we have the results from studies that em- ployed three different methods of estimation, we have used the approach of translating the results from the Lithuanian OE study and Swedish mail DC study into hypothetical DC face-to-face type of study results, using the Polish experience from all three types of surveys. In other words, the Polish main study results have been used as a common denominator in order to arrive at com- parable estimates. Based on Table 3 we assumed that the coefficient for translating an OE study results into a DC face-to-face study results equals 4, and for translating mail survey results equals 0.55. We have therefore assumed that a hypothetical DC face-to-face survey in Lithuania would have yielded WTP = 28, and in Sweden WTP = 252. To estimate the WTP for the entire Baltic re- gion we applied the following approach. We have assumed that the three countries that were subject of CVM surveys are representative for three sub-regions: Sweden for Western Europe market economies, Lithuania for the former So- viet Union republics, and Poland for Poland only. We assumed that within each of the groups of countries sub-regions WTP is proportional to GDP per capita at Purchasing Power Parity. Six extrapolations were thus made. The Swedish WTP was extrapolated to Finland, Denmark and Germany using ratios of the Swedish GDP and those of respective countries. Likewise, the Lithuanian WTP was extrapolated to Latvia, Es- tonia and Russia using ratios of the Lithuanian GDP and those of respective countries. The GDP figures and their ratios, b i , to the GDP of the reference country for each group are listed in Table 4. The same table contains WTP per cap- ita estimates based on the WTP attributed to a reference country and modified by the b i coeffi- cient. Aggregate WTP for countries, p i , are then calculated by multiplying mean WTP per capita by the number of adult population living in re- spective countries’ Baltic drainage basin Sweitzer et al., 1995; Rocznik, 1995. Table 4 Estimation of mean and total WTP values in countries of the Baltic Region a GDP per capita at Nominal GDP per Country b i, Mean WTP per Aggregate benefits p i capita US PPP US 10 6 US capita US 15 483 0.92 Finland 232 13 954 872 16 821 1 252 1615 Sweden 17 777 19 306 1.15 21 791 289 Denmark 997 19 688 Germany 18 541 1.10 278 676 4588 1 Poland 56 1911 1460 3632 1 573 28 Lithuania 73 765 Latvia 3058 0.84 24 46 Estonia 3823 956 1.05 29 33 4970 1.37 1147 38 Russia 276 7998 – 110 6048 Drainage Basin 6091 a Source of GDP data: OECD data files.

5. International cooperation